Fix instructions the model stopped following
Checks first whether a later request overrode an earlier rule, and says when to start over instead.
| Category | Using AI › Roles & personas |
|---|---|
| Tags | AnalyzingReviewingRewriting |
The model has stopped following my instructions as this conversation got long. **Work out why and fix it.** **Diagnose first:** 1. **Which instruction stopped being followed, and roughly when.** *Where in the conversation did the behavior change* 2. **Which of these it is:** - **The instruction was given once, early, and has scrolled far back** - **A later request contradicted it** — ***and the model chose the more recent one.*** **This is the most common cause** - The instruction was never concrete enough to check against - **Output from earlier turns is now acting as an example** — the model is imitating its own drift - The task changed enough that the instruction no longer fits 3. **Are any of my instructions mutually exclusive?** ***List them. I probably cannot see this myself*** **Then:** - **A restatement I can paste in now**, carrying only the instructions that still apply - ***Which instructions to drop*** — **keeping a dead rule alive is part of why the live ones stopped being followed** - **Whether to continue here or start fresh.** *If a fresh start is better, say so, and write the handoff*
After pasting, fill in the fields at the bottom (The conversation · What it stopped doing)
What this prompt does
The usual cause is not context length but a later request contradicting an earlier instruction — the model takes the recent one. This narrows the cause and finds your mutually exclusive rules.
More in this category
| Act as a Base LLM Model | |
| Act as a Conversational AI | |
| Aphorism Book | |
| Architectural Expert | |
| Artist Advisor |